首页|基于专利异构数据融合的技术演化路径识别方法

基于专利异构数据融合的技术演化路径识别方法

扫码查看
原文链接
  • 万方数据
  • 维普
[研究目的]针对目前技术演化分析中多关注专利文本,忽略专利引文信息的问题,提出一种基于专利异构数据融合的技术演化路径识别方法.[研究方法]首先,使用Sentence-BERT模型提取专利文本语义特征;其次,使用图卷积神经网络模型将文本语义特征与引文结构特征融合,实现异构数据融合构建专利向量;最后,划分时间窗,使用k-means算法对各时间窗进行技术主题聚类,基于相邻时间窗技术主题相似度构建技术演化路径.[研究结论]以人工智能领域为例进行实证研究,共发现4条技术演化路径.与相关权威报告进行比对,结果表明识别结果与人工智能技术领域的发展现状一致,验证了模型的有效性和科学性.
Technology Evolution Path Identification Method Based on Patent Heterogeneous Data Fusion
[Research purpose]Aiming at the current problem of focusing on patent text and ignoring patent citation information in tech-nology evolution analysis,a technology evolution path identification method based on patent heterogeneous data fusion is proposed.[Re-search method]First,the Sentence-BERT model is used to extract the semantic features of patent text;second,the graph convolutional neural network model is used to fuse the semantic features of text with the structural features of citations to realize the heterogeneous data fusion to construct the patent vectors;lastly,the time windows are divided,and the k-means algorithm is used to perform the technologi-cal theme clustering of the time windows,and the technological evolution paths are constructed based on the technological theme similarity of the neighboring time windows.[Research conclusion]Taking the field of artificial intelligence as an example for empirical research,a total of four technological evolution paths are identified.Comparison with relevant authoritative reports shows that the identification results are consistent with the development status quo in the field of artificial intelligence technology,which verifies the validity and scientific of the model.

patenttechnology evolutiontechnology evolution path identificationheterogeneous data fusionartificial intelligencesen-tence-BERTgraph convolutional neural network

侯艳辉、荆明月、王家坤

展开 >

山东科技大学经济管理学院 青岛 266590

专利 技术演化 技术演化路径识别 异构数据融合 人工智能 Sentence-BERT 图卷积神经网络

山东省自然科学基金项目

ZR2021QG035

2024

10.3969/j.issn.1002-1965.2024.09.023

情报杂志
陕西省科学技术信息研究所

情报杂志

CSTPCDCSSCICHSSCD北大核心
影响因子：1.502
ISSN：1002-1965
年,卷(期)：2024.43(9)
  • 20